Content-based Image Retrieval Research Under Network

With the rapid development of world and network, Multimedia informationincreases in explosion way. How to get the desired message in vast amounts ofMultimedia information becomes people’s urgent needs. Multimedia contains extensiveinformation, including text, image, video, and speech, etc. This paper is the departurefrom the multimedia category—images to explore the process of how the network canobtain the desired images.Firstly, this thesis is guided by the contents of image feature, respectively, form theimage’s color, image’s sharp and texture three aspects of feature extraction. Colorfeatures mainly research on the Color Histogram and analyze the Color Moment, ColorCoherence Vector and Color Correlation Chart. Sharp features are expanded from theimage region, border and Skeleton. Region studies include Invariant Moments andFourier operator. Border mainly researches on the Direction Chain Code. Skeleton doessome analysis of Voronoi and Reeb Graph. Texture mainly researches on the JointProbability Matrix and the LBP algorithm. Secondly, after gotten the image features,how to optimally process it by the computer is the key, including optimization ofsubjective characteristics, PCA feature optimization and code optimization feature.Thirdly, in order to obtain high-level semantics of the image, writer establishes a simplesemantic classification models and analyzes the way of establishment of image semanticlibrary by image related text. Finally, by introducing the realization retrieve schematics,function modules, related technologies and combined related algorithms, writerpractices the image content retrieval and does some comparison of single feature-basedsearch and various features-based result. The corresponding conclusions and analysissome places for improvement are given in the final.